BB [[144,12,12]] on QCCD — Bravyi's syndrome round, every design ranked

What is compiled. Bravyi et al.'s depth-8 round (Fig. 5 of arXiv:2308.07915) as published: one ancilla per check, so 144 data ions and 144 check ions, 288 in all, and its 864 CNOTs in the schedule's order. Every programme here was compiled from that circuit by the released compiler, qccdc cd3efedbbc10, which keeps each qubit's gates in the circuit's order; the checker confirms it on 8 of 8: no operation starts before the one before it on the same qubit has finished. Cooling is inserted where the heating model needs it, the one change the release allows, and R10 is decided by the proved Lean checker and the stabilizer tableau (passed on 8 of 8).

What is ranked. 7 designs whose programme passes every rule on the replay and whose machine fits the task's hardware budget: loops without docks whose traps hold four ions, which the general router serves; rings with a dock for every check, which rigid rotation serves. The fastest round on the jones table: ring(100,2,0) 4-ion traps, 288.4 ms; among the ring + docks designs, ring(145,3,144) junction and ring(146,2,144) junction, 1,000.8 ms. Pick a measure to rank by, and click a dot for its numbers and the steps where each claim can be checked. 1 more design runs the round but breaks a rule or needs more hardware than the task allows: listed under Disqualified, below, never ranked.

What does not compile yet. grid13x13, ring(146,2,72) junction, ring(76,2,0) 4-ion traps, racetrack(150): our compiler does not compile them within the release's 30 minutes, and each one’s page gives the compiler's own reason. They are listed as refused, at the right of the plot, and never ranked; the router work for these shapes is a separate follow-up.